Achievement of dynamic tennis swing motion by offline motion planning and online trajectory modification based on optimization with a humanoid robot

Abstract

In order for a humanoid robot to achieve higher physical performance, it is important to generate dynamic whole-body motion under the constraints of physical limitations and dynamic balance. In this paper, we propose the method for generating dynamic whole-body motion such as sports motion based on optimization techniques. Taking a tennis forehand swing as an example of dynamic motion, we aim to increase the swing speed. The proposed methods are composed of the following two methods: 1) the offline swing optimization and 2) the online swing modification. In the offline swing optimization, we use non-linear optimization techniques to maximize the swing speed and satisfy the constraints. We generate all joint trajectories by optimizing control points of uniform B-splines. In the online swing modification, we modify a part of the optimized trajectories online considering joint velocity limits, since the predicted ball trajectory might change before a robot hits the ball. These methods are validated through the following experiments. First, we carry out the experiment in which the actual robot JAXON executes the optimal swing motion. We confirm that the method of the offline swing optimization generates the feasible motion which reaches up to 14.6m/s. We also apply the online swing modification to the optimized motion in the simulation. Then we evaluate how accurately we have to predict the ball trajectory.